Pro-softpatent analysts have yet to find benefit from software patents.
Rather than presenting a one-sided story, let us begin with the scholarship by pro-softpatent academics. Many respected academics have sought empirical evidence that software patents foster innovation. What have they found?
- Ronald Mann interviewed venture capitalists and programmers, searching for those who would state that they rely on patents to evaluate a company, but found that all of his interviewees largely just ignored software patents. 
- On a quantitative scale, Mann and Sager searched a venture capital database for evidence that software patents help startups. Their conclusion: "…[W]e found no significance to the existence of pre-financing patents.'' 
- Robert Merges is the coauthor of the standard text on digital intellectual property . His endorsements have been tepid at best: "I will venture a prediction: Patents will not cause any real and lasting problems. [p 12] … Something good may come of it. [p 13]'' 
- Merges also produced an empirical paper that again was unable to find evidence to back up any claims that patents foster innovation. After the full empirical analysis, Merges was left with the same endorsement for software patents as before: "Whatever the effects of patents on the software industry, this paper concludes, they have not killed it." 
- Lerner and Zhu attempted a similar search for empirical evidence of benefits from software patents, and were also left with an optimistic spin on null results: "… little evidence can be found for harmful effects…." 
In this article, Jim Bessen questions the validity of these null-result findings. After all, statistical theory is built with a bias toward skepticism: there is no statistical test that will accept a null hypothesis as true. A null result could mean that the researcher is looking in the wrong place, has misspecified the model, or is using an underpowered test.
In fact, Cockburn and MacGarvie  used more careful empirical methods, focusing on subindustries within the broad category of software, and did find a stifling effect to software patents: "Start-up software companies operating in markets characterized by denser patent thickets see their initial acquisition of VC funding delayed relative to firms in markets less affected by patents."
Software patents affect more than just software companies.
The latest accounting from the Bureau of Economic Analysis divides the software market into three parts: retail, consultants, and in-house, which are evenly split in the U.S. economy. Of the $232.5 billion spent on software in 2002, 32.6 percent bought prepackaged programs, 36.4 percent custom-built ones, and 31.0 percent paid for software written in-house. A report commissioned by the EU used a different method and different definitions that indicate that software sales represents an even smaller portion of the software economy: by their estimates, 16% of 2002 U.S. software sales went to proprietary software sales, 41% went to development and customization services, and 43% paid for internal development. 
Many of the patent infringement suits that we see, such as the infamous One-click case of Amazon v Barnes and Noble, or Acacia Media Technologies' suits against Johns Hopkins University, the University of Virginia, the University of Wyoming, and a number of other schools , or Global Patent Holdings v Green Bay Packers, consist of a software-oriented company suing a company that is not necessarily software-oriented, but does have a web site.
Therefore, to search for the effect of software patents on software companies is to look at, at best, a third of software users—and the third that is most likely to benefit from a government-granted monopoly.
Software is a complex industry.
Intuitively, one would expect that the transactions costs caused by patents would increase as more patents bear on a single output. For example, most drugs are covered by one or two patents, while computer hardware and software is covered by potentially thousands of patents. This argument is formalized in Bessen and Maskin .
Does this argue that all complex industries should be outside the scope of patent law, not just software? Perhaps by itself, but we should also consider the context in which software patents are used, which is different from how hardware patents are used. Cohen, Nelson, and Walsh  surveyed patent holders as to why they hold patents, and found that those in the computer industry were much more likely to hold patents for defense and other horse-trading relative to those in other complex fields such as semiconductors; members of other complex fields were more likely to patent to protect their invention, as patents are generally understood to be intended.
Patent suits cost billions of dollars per year.
Based on estimates by Bessen and Meurer, we estimate costs from software patent lawsuits of $11.2 billion. [See our 2008 State of Softpatents report for details.]
Further, this considers only the litigation that goes to court. Kahin also cites the "Rule of 25" from Chip Lutton of Apple Corporation: for every case filed, there are 25 that are settled before that stage; for every case that is settled, there are 25 patent infringement letters. Every letter sent could cost up to $50,000 for the required legal opinion. It is impossible to determine the exact cost caused by infringement letters, because there is no easy way to count them, but their costs indicate that the estimate of costs of $11.2 billion from litigation filed with the courts is a low estimate of the costs.
Again, notice that many of the costs in both litigation and stifled activity are borne by defendants outside of the narrowly-defined software industry, and therefore would have been ignored by the above empirical investigations.
Government intervention in the market is generally taken to be a last resort
Simply put, a patent is a government-granted monopoly. Of course, there are many reasons why a government would intervene with the free market, and protecting a research investment is one of them. The modern pharmaceutical industry has never existed without patent protection, and it is an open question (far beyond the scope of the ESP) as to whether it could exist without government support. But software innovation has occured without patent protection, in spades. Word processors, spreadsheets, databases, email, the World Wide Web, audio-video compression/decompression software, the fundamentals of all modern operating systems, and the basic elements of the graphical user interface were all developed before software was widely understood to be patentable, and was thus developed without patent support. Modern advances, like RSS, XML, and the AJAX standards, were all developed and popularized without patents. As noted by the pro-patent empiricists above, patents are irrelevant to advancement in software. So why is the government providing this type of selective and distortionary monopoly, which causes private parties to spend billions of dollars in legal fees?
- J Bessen and R Hunt, An Empirical Look at Software Patents, Federal Reserve Board of Philadelphia Working Paper No. 03-17.
- J Bessen and M Meurer, Patent Failure: How Judges, Bureaucrats, and Lawyers Put Innovators at Risk. Forthcoming, Princeton University Press. See especially chapter 9: "Abstract patents and software". [Preview page.]
- B Klemens, Math You Can't Use: Patents, Copyright, and Software, Brookings Institution Press, 2005. Brookings link